Data Science Seminar

The LTCI Data Science Seminar is a joint seminar between the DBWeb team and the S2A team of the LTCI. It features talks related to machine learning and data management.

Attendance is generally open to the public, feel free to contact us if you are interested in coming. Talks are held at Télécom ParisTech, 46 rue Barrault, Paris, France, métro Corvisart.

September 7, 2017

The seminar takes place from 2PM to 4PM in Amphi Saphir, and consists of the two following talks:

Talk 1: Albert Bifet
Massive Online Analytics for the Internet of Things (IoT)

Abstract: Big Data and the Internet of Things (IoT) have the potential
to fundamentally shift the way we interact with our surroundings. The
challenge of deriving insights from the Internet of Things (IoT) has
been recognized as one of the most exciting and key opportunities for
both academia and industry. Advanced analysis of big data streams from
sensors and devices is bound to become a key area of data mining
research as the number of applications requiring such processing
increases. Dealing with the evolution over time of such data streams,
i.e., with concepts that drift or change completely, is one of the
core issues in stream mining. In this talk, I will present an
overview of data stream mining, and I will introduce
some popular open source tools for data stream mining.

Talk 2: François Roueff,
Prediction of weakly locally stationary processes by auto-regression

Abstract: We introduce locally stationary time series through the  local approximation
of the non-stationary covariance structure by a  stationary one. This allows
us to define autoregression coefficients in a  non-stationary context, which,
in the particular case of a locally stationary  Time Varying Autoregressive
(TVAR) process, coincide with the generating coefficients. We provide and
study an estimator of the time varying autoregression coefficients in a
general setting. The proposed estimator of  these coefficients enjoys an
optimal minimax convergence rate under limited  smoothness conditions.
In a second step, using a bias reduction technique, we  derive a minimax-rate
estimator for arbitrarily smooth time-evolving  coefficients, which outperforms
the previous one for large data sets. In  turn, for TVAR processes, the predictor
derived from the estimator  exhibits an optimal minimax prediction rate.

 

June 22, 2017

The seminar takes place from 2PM to 4PM in Amphi Saphir, and consists of the two following talks:

Talk 1: Pascal Bianchi,
Distributed optimization on graphs using operator splitting methods

You can download the slides of this talk.

Abstract: Consider a network of N agents (computing units) having private
objective functions and seeking to find a minimum of the aggregate
objective. The aim is to design iterative algorithms where, at a each
iteration, an agent updates a local estimate of the minimizer based on the
sole knowledge of its private function and the information received from
its neighbors. In this talk, i will first provide an overview of standard
distributed optimization methods. Then, i will explain how recent and
generic results in stochastic optimization can be used in order to design
asynchronous and adaptive distributed optimization algorithms.

Talk 2: Maximilien Danisch,
Towards real-world graph algorithmics

You can download the slides of this talk.

Abstract: Real-world graphs (a.k.a. complex networks) are ubiquitous: the web, Facebook,
brain networks, protein interaction networks, etc. Studying these graphs and
solving computational problems on them (say maximum clique, partitioning or
dense subgraph) has applications in many fields. I will first show that the
structure of some real-world graphs is special. In particular, they are not
adversarial and some difficult problems (say NP-hard problems) can be solved
on some huge real-world graphs (say 2G edges or more). I will then present two
works along the lines of “understanding and leveraging the structure of
real-world graphs in order to build better algorithms”: (i) Enumerating all
k-cliques and (ii) Computing the density-friendly graph decomposition. The
latter one has been published in WWW2017.